[Samba] Samba hangs badly making the OS hang unrecoverably too

[Alberto Di Fede]

Hi everybody,

i'm getting mad regarding a very serious (since it's happening on a
production server) issue with a samba domain joined installation.

This machine has worked perfectly for an year and the last configuration
tune up and version upgrade (3.0.23c) went along fine in the first days of
January.

It's been two weeks approximately now since we started to have a bad issue,
a smbd lock causing a complete machine lock.

I don't wont to get too much into annoying details, but let's say that an
hardware problem has been ruled out, that there are no network
communications issues and that, most interestingly, we had the same issue
with another identically configured server.

In my young sysadmin career (6 years now) i've never seen such a failure,
one causing a linux machine to stop responding on console commands (even ps
aux, lsof...), never letting me to look at logs while on hang (console will
hang waiting for the file open), neither to kill processes hanging.
There's no CPU occupation, nor particular network activity neither resources
depleting: it just hangs there with no possibilities to recover it besides
an hardware machine reset.

All of this considered, i tried to debug the problem in every way, looking
for particular messages into every log and launching every samba related
daemon interactively and with debug level at 10: i obtained no particular
info, no write or read error, communications problems with the domain of any
nature. NMBD and SMBD processes just stop to work and hangs waiting, i
really don't know for what.

When it hangs i'm still able to communicate with open ssh sessions as long
as i i don't try to look for open processes or files. I suppose is something
related to name resolution and winbind rid mappings, but from what i've
gathered so far, i've no particular info to spot even possibly the problem.

Any help or suggestion from you would be very welcomed and appreciated.

Tomorrow i will setup another machine for the same purpose with an opensuse
10.2 and look what happens.

Thanks a lot in advance to everybody willing to help, i'll be happy to
provide you with any requested further info.


Alberto


This is my smb.conf

[global]
   workgroup = AGBSOFT
   realm = AGBSOFT.CH
   netbios name = FTP
   server string = FTP Server
   wins server = 10.100.0.2,10.100.0.4
   #client schannel = no
   idmap uid = 10000-200000000
   idmap gid = 10000-200000000
   idmap backend = rid:AGBSOFT=10000-200000000
   allow trusted domains = no
   winbind enum users = yes
   winbind enum groups = yes
   winbind use default domain = yes
   winbind nested groups = yes
   template homedir = /home/%D/%U
   template shell = /bin/bash
   load printers = no
   log file = /var/log/samba/%m.log
   #log level = 8
   max log size = 0
   security = ads
   #password server = agbsoft-nt1.agbsoft.ch
   encrypt passwords = yes
   socket options = IPTOS_LOWDELAY TCP_NODELAY
   #os level = 23
   domain master = no
   preferred master = no
   local master = no
   inherit acls = yes
   inherit permissions = yes
   map acl inherit = yes
   store dos attributes = yes
   acl compatibility = win2k
   #acl group control = yes
   map hidden = no
   map system = no
   map readonly = no
   nt acl support = yes
   ea support = yes
   winbind offline logon = true
   winbind refresh tickets = true
   dos filemode = yes

[FTPSpace]
   comment = "FTP"
   path = /ftp
   valid users = @"AGBSOFT\Domain Users"
   writable = yes

[Regressions]
   comment = "Regressions Files"
   path= /ftp/istap/Regressions
   valid users = @"AGBSOFT\Domain Users"
   writable = yes